Related Experiment Video
Updated: Jul 1, 2026

Establishing a Competing Risk Regression Nomogram Model for Survival Data
Published on: October 23, 2020
Predicting long-term absolute risk of lung cancer in patients with Hodgkin lymphoma
Shahin Roshani1, Naomi B Boekel1, Flora E Van Leeuwen1
1Department of Epidemiology, Netherlands Cancer Institute, Amsterdam, The Netherlands.
Background:
Patients with Hodgkin lymphoma are at increased risk of developing lung cancer, especially after chest radiation therapy. No tools are currently available, however, to predict lung cancer risk for different Hodgkin lymphoma treatments.
Methods:
In a cohort of 5370 individuals who had survived Hodgkin lymphoma for at least 5 years, were 15 to 50 years of age at the time of Hodgkin lymphoma diagnosis, and who were treated in the Netherlands between 1965 and 2012, we used radiation therapy fields and prescribed dose to estimate mean lung dose. The twinning method was used to divide the cohort into homogenous parts: 80% for model development and 20% for validation using inverse probability of censoring weighting area under the curve. Cox proportional hazards models allowing for time-dependent coefficients were used to model time from Hodgkin lymphoma diagnosis to lung cancer and lung cancer-free death as the competing event for predicting absolute lung cancer risk up to 30 years after Hodgkin lymphoma diagnosis.
Results:
Treatment information consisted of supradiaphragmatic radiation therapy in 75.2% of patients with a mean lung dose of 15.8 Gy. During follow-up, 218 survivors developed lung cancer. Older age, male sex, smoking at the time of Hodgkin lymphoma diagnosis, and higher mean lung dose were associated with higher lung cancer risk. The median Inter Quartile Range (IQR) estimated 30-year absolute lung cancer risk in our cohort was 1.2% (0.8%-1.9%) in nonsmokers and 6.9% (4.6%-10.9%) in smokers at the time of Hodgkin lymphoma diagnosis. The 20-year and 30-year inverse probability of censoring weighting areas under the curve values were 0.79 (95% Confidence Interval (CI) = 0.73 to 0.85) and 0.75 (95% CI = 0.69 to 0.81), respectively.
Conclusion:
We developed a well-calibrated prediction tool that estimates long-term risk of lung cancer with good discrimination based on patient characteristics and mean lung dose, allowing application in Hodgkin lymphoma survivors and contemporary patients with Hodgkin lymphoma.
Related Concept Videos
Statistical Methods for Analyzing Epidemiological Data
Cancer Survival Analysis
